Exporters which use probing to collect metrics (such as blackbox-exporter or snmp-exporter) always return up metric with value true as long as exporter itself is running. The target API of Prometheus also returns health: up for such instances, regardless if the instance being probed is not reachable. In such cases a metric probe_success should be considered to determine the up status of target.
Alertmonitor will incorrectly consider such targets as Up, meaning that at least one exporter on this target/instance is reachable (eg. blackbox-exporter). But it is not.
Exporters which use probing to collect metrics (such as blackbox-exporter or snmp-exporter) always return
upmetric with valuetrueas long as exporter itself is running. ThetargetAPI of Prometheus also returnshealth: upfor such instances, regardless if the instance being probed is not reachable. In such cases a metricprobe_successshould be considered to determine the up status of target.Alertmonitor will incorrectly consider such targets as
Up, meaning that at least one exporter on this target/instance is reachable (eg. blackbox-exporter). But it is not.